

Captum · Model Interpretability for PyTorch
Interpret models in PyTorch
- Website
- captum.ai
- Category
- Code & IT › AI Code Assistant
- Pricing
- Open source
- Platforms
- website
About Code & IT tools
Code & IT tools help you accomplish a range of AI-powered tasks. Captum · Model Interpretability for PyTorch is one option in this category.
About Captum · Model Interpretability for PyTorch
Supports interpretability of models across modalities including vision, text, and more. Supports most types of PyTorch models and can be used with minimal modification to the original neural network. Open source, generic library for interpretability research. Easily implement and benchmark new algorithms. Install Captum: via conda (recommended): conda install captum -c pytorch via pip: pip install captum Create and prepare model: import numpy as np import torch import torch.nn as nn from captum.attr import IntegratedGradients class ToyModel (nn.Module) : def __init__ (self) : super().__init__() self.lin1 = nn.Linear( 3 , 3 ) self.relu = nn.ReLU() self.lin2 = nn.Linear( 3 , 2 ) # initialize weights and biases self.lin1.weight = nn.Parameter(torch.arange( -4.0 , 5.0 ).view( 3 , 3 )) self.lin1.bias = nn.Parameter(torch.zeros( 1 , 3 )) self.lin2.weight = nn.Parameter(torch.arange( -3.0 , 3.0 ).view( 2 , 3 )) self.lin2.bias = nn.Parameter(torch.ones( 1 , 2 )) def forward (self, input) : return self.lin2(self.relu(self.lin1(input))) model = ToyModel() model.eval()
Features
Install Captum: via conda (recommended): conda install captum -c pytorch via pip: pip install captum
To make computations deterministic, let's fix random seeds: torch.manual_seed( 123 ) np.random.seed( 123 )
Define input and baseline tensors: input = torch.rand( 2 , 3 ) baseline = torch.zeros( 2 , 3 )
View Output: IG Attributions: tensor([[ -0.5922 , -1.5497 , -1.0067 ], [ 0.0000 , -0.2219 , -5.1991 ]]) Convergence Delta: tensor([ 2.3842e-07 , -4.7684e-07 ])
Use cases
Use case 1: Install Captum: via conda (recommended): conda install captum -c pytorch via pip: pip install captum
Use this capability in a real workflow, not only as a demo.
Use case 2: To make computations deterministic, let's fix random seeds: torch.manual_seed( 123 ) np.random.seed( 123 )
Use this capability in a real workflow, not only as a demo.
Use case 3: Define input and baseline tensors: input = torch.rand( 2 , 3 ) baseline = torch.zeros( 2 , 3 )
Use this capability in a real workflow, not only as a demo.
Getting started
- Visit Captum · Model Interpretability for PyTorch's website to create an account or start using it
- Available on: website
- Check the pricing page for the plan that fits
Screenshots

Related tools
Compare Captum · Model Interpretability for PyTorch with other Code & IT tools such as ModelsLab, Music.AI before you pick one.




